280 Chem. Res. Toxicol. 2001, 14, 280-294

Structure-Activity Relationships for a Large Diverse Set of Natural, Synthetic, and Environmental

Hong Fang,† Weida Tong,*,† Leming M. Shi,†,‡ Robert Blair,§ Roger Perkins,† William Branham,§ Bruce S. Hass,§ Qian Xie,† Stacy L. Dial,§ Carrie L. Moland,§ and Daniel M. Sheehan§

R.O.W. Sciences, Inc., 3900 NCTR Road, MC 910, Jefferson, Arkansas 72079, and Division of Genetic and Reproductive Toxicology, National Center for Toxicological Research, Jefferson, Arkansas 72079 Received October 3, 2000

Understanding structural requirements for a chemical to exhibit receptor (ER) binding has been important in various fields. This knowledge has been directly and indirectly applied to design drugs for human estrogen replacement therapy, and to identify estrogenic endocrine disruptors. This paper reports structure-activity relationships (SARs) based on a total of 230 chemicals, including both natural and . Activities were generated using a validated ER competitive binding assay, which covers a 106-fold range. This study is focused on identification of structural commonalities among diverse ER ligands. It provides an overall picture of how xenoestrogens structurally resemble endogenous 17â- (E2) and the synthetic estrogen (DES). On the basis of SAR analysis, five distinguishing criteria were found to be essential for activity, using E2 as a template: (1) H-bonding ability of the phenolic ring mimicking the 3-OH, (2) H-bond donor mimicking the17â-OH and O-O distance between 3- and 17â-OH, (3) precise steric hydrophobic centers mimicking steric 7R- and 11â-substituents, (4) hydrophobicity, and (5) a ring structure. The 3-position H-bonding ability of phenols is a significant requirement for ER binding. This contributes as both a H-bond donor and acceptor, although predominantly as a donor. However, the 17â-OH contributes as a H-bond donor only. The precise space (the size and orientation) of steric hydrophobic bulk groups is as important as a 17â-OH. Where a direct comparison can be made, strong estrogens tend to be more hydrophobic. A rigid ring structure favors ER binding. The knowledge derived from this study is rationalized into a set of hierarchical rules that will be useful in guidance for identification of potential estrogens.

Introduction turally diverse chemicals have estrogenic activity like the There is a growing body of evidence that some man- endogenous hormone estradiol. made chemicals, now called endocrine-disrupting chemi- SARs for estrogens date back more than six decades cals (EDCs),1 have the potential to disrupt the endocrine to the early work of Dodds et al. (4, 5). The succeeding system by mimicking endogenous hormones such as two decades saw the discovery of estrogens, estrogens and androgens (1). Recent legislation mandates such as DES, based on understanding of the important that the Environmental Protection Agency develop a structural features governing potency for steroidal es- screening and testing program for potential EDCs, of trogens. Recently, a number of SAR studies have been which xenoestrogens figure predominately (2). Xenoestro- reported for steroidal estrogens (6) and nonsteroidal gens contain a number of chemical classes that display estrogens (7). These are generally focused on identifica- a broad range of structural diversity (3). For example, tion of structural characteristics for chemicals within DES, , polychlorinated biphenyls (PCBs), alkylphe- similar two-dimensional (2D) structural frameworks, nols, phthalates, and have been found to be including E2 derivatives (6), DES derivatives (8), PCBs estrogenic. It has long been an enigma why such struc- (9), (10), alkylphenols (11), raloxifenes (12), and others. Modern computer-based tools have * To whom correspondence should be addressed: R.O.W. Sciences, enabled the development of quantitative structure- Inc., 3900 NCTR Road, MC 910, Jefferson, AR 72079. Telephone: (870) 543-7142. Fax: (870) 543-7382. E-mail: [email protected]. activity relationship (QSAR) models for identifying steric † R.O.W. Sciences, Inc. and electrostatic features of a molecule in three-dimen- ‡ Current address: BASF Corp., P.O. Box 400, Princeton, NJ 08543. sional (3D) space for estrogenic activity (8, 13-19). § National Center for Toxicological Research. 1 Abbreviations: SAR, structure-activity relationship; QSAR, quan- Recent crystallographic structures of the human ER R - R titative structure activity relationship; ER, ; hER , subtype (hERR) with a number of ligands, including E2, human estrogen receptor R subtype; EDCs, endocrine disrupting chemicals; RBA, relative binding affinity; NA, not active; NCTR, DES, , and 4-OH-tamoxifene, have also been National Center for Toxicological Research; 2D, two-dimensional; 3D, reported (20, 21). By aligning these four ligands on the three-dimensional; E2,17â-estradiol; E1, ; E3, ; EE, basis of the superposition of their ER binding sites, we ethynylestradiol; DES, diethylstilbestrol; DMS, dimethylstilbestrol; DDTs, 1,1,1-trichloro-2,2′-bischlorophenylethane derivatives; PCBs, have been able to demonstrate the common binding polychlorinated biphenyls; log P, hydrophobicity. characteristics among these ligands (22).

10.1021/tx000208y CCC: $20.00 © 2001 American Chemical Society Published on Web 02/10/2001 SARs of Diverse Estrogens Chem. Res. Toxicol., Vol. 14, No. 3, 2001 281

In principle, chemicals with similar biological activity and multiplying by 100 (E2 RBA ) 100). The validated assay share common structural features. This implies that incubation conditions were 20 h at 4 °C using 17 mg of uterine ) 3 structurally diverse estrogens possess a certain degree tissue/mL (Bmax 0.22 nM) with 1 nM [ H]E2. The competing of structural commonality essential to eliciting estrogenic chemical concentrations ranged from 1 nM to 1 mM. Chemicals 3 activity. Although a number of chemical classes are that failed to compete for [ H]E2 binding to the ER were designated as “not active” (NA). Chemicals that exhibited known to be estrogenic, little thorough structural evalu- binding, but did not reach 50% inhibition in the designed ation has been presented for their structural resemblance concentration range, were designated as “slight binders”. All to the E2 or the strong synthetic estrogen DES. Moreover, assays were repeated at least twice; the IC50 values of positive only limited efforts have explored the structural similari- chemicals are the means of the replicate values. The standard ties between different chemical classes of xenoestrogens. deviation of IC50 for each chemical was reported (23), and only To better understand the structural requirements for ER the mean RBA value was used for this study. The purity of binding, it is important to have a reliable data set, chemicals as well as their effect on RBAs was also studied by obtained with consistent assay design, covering a broad Blair et al. (23). The largest fold difference (∼10-fold) was found range of chemical classes. Recently, we reported ER for from different commercial sources due to the binding activity data for a large number of chemicals, impurity of the sample. including natural, synthetic, and environmental estro- Molecular Modeling. The crystal structures of E2, DES, raloxifene, and 4-OH- bound to the ER were obtained gens, using a validated rat ER competitive binding assay from the Protein Data Bank (PDB) as entries 1A52, 3ERD, (23, 24). This data set, called the NCTR data set, 1ERR, and 3ERT, respectively. The alignment of these four currently contains 230 chemicals. It was designed a priori ligands, based on root-mean-square (RMS) fitting of their to cover broad structural diversity and a wide range of receptor coordinates, was performed using the InsightII software binding activities for elucidating the structural charac- package (Molecular Simulations, Inc., San Diego, CA). Log P teristics of xenoestrogens and natural estrogens. was calculated using the atom/fragment contribution method The rat uterine cytosol ER competitive binding assay (27). Pharmacophore searching was performed with the CATA- is the gold standard for in vitro ER assays. When our LYST package (Molecular Simulations, Inc.). The energy dif- results are compared to results from other ER binding ferences for E2 and DES between the conformation in their binding modes and that in the minimum conformation of the assays, there is a general consistency between relative free ligands were calculated using the AM1 model Hamiltonian ER activities across different assay methods and species of the AMPAC/MOPAC module in InsightII (Molecular Simula- (25). For example, we found a high linear correlation for tions, Inc.). The atom-atom distance was also measured using ER binding affinities among a diverse group of chemicals InsightII (Molecular Simulations, Inc.). assayed with the ER from rat uterine cytosol and hERR.2 Further, we also found that ER assay results correlated Results very well with those from a yeast-based reporter gene assay and a MCF-7 cell proliferation assay. These find- We determined the ER RBA of 230 chemicals, of which ings demonstrate that ER binding is the major determi- 130 were active and 100 inactive. To the best of our nant across three levels of biological complexity (receptor knowledge, this is the largest published ER competitive binding, a yeast reporter gene response, and cell prolif- binding data set. This NCTR data set has been exten- eration) of estrogen action. Moreover, chemicals positive sively used to build and validate a series of computational in uterotrophic responses (in vivo estrogenic activity) are models proposed for priority setting of potential estro- also positive in the ER binding assay, indicating that genic EDCs (22). binding affinity is a good predictor of in vivo activity with For the convenience of analysis and presentation, the few false negatives observed (26). Therefore, understand- NCTR data set is divided into seven major categories ing the structural requirements for ER binding provides according to the chemical structural characteristics. The strong guidance in identifying potential in vivo estrogenic chemical names and structures, as well as RBA values, EDCs. are shown in Figures 1-7 for 130 active chemicals and The NCTR data set covers most known estrogenic a few of the selected inactive chemicals. are chemical classes as well as some new estrogenic chemi- chemicals with a steroidal backbone. E2 derivatives have cals. A careful SAR examination of the NCTR data set a phenolic A ring (Figure 1A), whereas the others lack in conjunction with knowledge of the recently reported a phenolic A ring (Figure 1B). The common structural ligand-ER crystal structures should allow a better feature of DES-like chemicals is two benzene rings sepa- understanding of the general structural requirements of rated by two carbons that are connected by a double bond a chemical binding to the ER in a quantitative manner. in DES derivatives (Figure 2A) and by a single bond in In turn, the knowledge can be used to develop a predic- derivatives (Figure 2B). In this group of chemi- tive toxicology model for rapidly identifying potential cals, derivatives (Figure 2C) are struc- estrogenic EDCs. turally similar to DES derivatives, but have an additional phenyl group attached to the ethylene bridge group. Most Materials and Methods synthetic contain this structural feature. Phytoestrogens contain four major structurally distinct ER Binding Activities. The ER binding affinities of chemi- chemical classes: , , , cals were determined by using a competitive receptor binding assay described previously (23). Briefly, a chemical’s binding and . Flavonoids are the largest class, activity was determined by competing with radiolabeled [3H]- containing flavones (Figure 3A), flavanones (Figure 3B), E2 for the ER in rat uterine cytosol. The IC50 (50% inhibition of and isoflavones (Figure 3C). They share a benzene ring 3 [ H]E2 binding) for each competitor was determined. The directly connected to a chromone or 4-chromanone. relative binding affinity (RBA) for each competitor was calcu- Coumestans (Figure 3D) contain the smallest number of lated by dividing the IC50 of E2 by the IC50 of the competitor chemicals, of which is the most rigid and flat molecule of all the phytoestrogens used. In contrast, 2 H. Fang, W. Tong, and D. Sheehan, unpublished results. chalconoids (Figure 3E) are the most structurally flexible 282 Chem. Res. Toxicol., Vol. 14, No. 3, 2001 Fang et al.

Figure 1. Steroids with (A) and without (B) a phenolic ring. chemicals of the phytoestrogens. The generic structure chemical classes. Table 1 shows the mean RBA values of chalconoids is two benzene rings separated by three for each chemical class. Steroids and DES-like chemicals carbons. Mycoestrogens (Figure 3F) have a large ring have the strongest binding activities. The RBAs of fused with a benzene ring that contains two OH groups the rest of the chemical classes follow this order: phyto- at positions 14 and 16. Chemicals with two benzene rings estrogens > diphenylmethanes > biphenyls > phenols. connected by one carbon atom are classified as diphe- There are only a few inactive chemicals in the DES-like nylmethanes, including diphenolalkanes (Figure 4A), chemical and phenol classes. The mean RBA of DES-like benzophenones (Figure 4B), and DDTs (Figure 4C). chemicals is 2.63, which is the highest of the groups that Biphenyls are chemicals with two phenyl rings directly were examined, while phenols (mean RBA ) 0.0015) has connected to each other. They can be chlorinated as PCBs the lowest mean RBA. These observations suggest that (Figure 5A) or nonchlorinated (Figure 5B). Alkylphenols although both classes are likely to bind to the ER, their (Figure 6A), parabens (Figure 6B), and alkyloxyphenols binding specificity is different due to the nature of their (Figure 6C) are categorized as phenols. They all contain structures. a single phenolic ring. Most chemicals in this group have Steroids. E2, with an RBA of 100, is one of the most a long alkyl chain substituted at the para position. The active estrogens. It has a hydrophobic backbone with OH chemicals that do not belong to any of these groups are groups at each end of the molecule. Such a molecular found in miscellaneous (Figure 7). configuration was considered as the most important Chemicals that exhibit ER binding have a broad range structural features for ER binding (6). The recently of structural diversity and RBAs. A common structural reported crystal structure of the ER-E2 complex reveals feature for steroids, DES-like chemicals, and most phyto- that the 3- and 17â-OH groups primarily serve as H-bond estrogens is the presence of two rings (one of them donors and acceptors in interacting with the receptor usually a phenolic ring) separated by two carbons. binding site (20). The elimination or modification of either Chemicals either with two rings separated by one carbon of these two OH groups significantly reduces a chemical’s atom (diphenylmethanes), connected directly (PCBs), or binding affinity for the receptor, as shown in Table 2. possessing only one ring (alkylphenols, phthalates, and This impact is more dramatic at the 3-position than at kepone) typically have relatively lower binding affinities the 17â-position. For example, the loss of binding activity than chemicals with two rings separated by two atoms of 3-deoxy-E2 is 28-fold greater than the loss for 17-deoxy- (25), although there is overlap in RBAs among some E2, indicating that the 3-OH is more important than the SARs of Diverse Estrogens Chem. Res. Toxicol., Vol. 14, No. 3, 2001 283

Table 1. General Information for the ER Ligand Categories, Including the Number of Chemicals with the Ratio of Active and Inactive Ligands, Mean RBA Values,a and Representative Chemicals with Key Structural Featuresb

a Mean RBA values are calculated for active ligands only. b The presence and absence of a structural feature are represented as 1 and c 0, respectively. A ring mimics one of the B, C, and D rings of E2.

Table 2. Effects of Elimination and Modification of the (RBA ) 0.002, dO-O ) 9.95 Å) is 60-fold lower than that 3- and 17â-OH on RBAs of its â-isomer. elimination or from to fold The RBA of 17-deoxy-E2 (RBA ) 14.1) was slightly modification (RBA) (RBA) difference larger than that of estrone (RBA ) 7.31), indicating that 3-OH E2 (100) 3-deoxy-E2 (0.5) 200 the 17-ketone group might not contribute significantly E1 (7.31) 3-deoxy-E1 (0.006) 1218 to the binding as a H-bond acceptor. If the ketone group EE (190) (2.26) 84 of E was moved from the 17-position to the 16-position E (9.72) 3-methylestriol 442 1 3 to form estra-1,3,5(10)-trien-16-one-3-ol, the RBA was 17â-OH E2 (100) 17-deoxy-E2 (14.1) 7 E2 (100) E1 (7.31) 14 decreased 16-fold. A similar activity reduction was also observed with introduction of an OH group at the 16R- 17â-OH in ER binding. This is consistent with the position of E2, resulting in a relatively weak estrogen, E . This demonstrates that a small polar group at the findings from the ER-E2 crystal structure that the 3-OH 3 has H-bonding interactions with Glu 353, Arg 394, and 16-position reduces a chemical’s activity. a water molecule, whereas 17â-OH only forms one Systematic studies on the influences of substituents H-bond with His 524. at various positions of E2 revealed that for most positions The elimination of 3-OH caused a greater reduction the introduction of substituents results in a loss of in the activity for estrone (E1) than for E2 (Table 2). A binding affinity (17, 28). For example, the E2 metabolites similar result was also observed for 3-OH methylation 2-OH-E2 (RBA ) 22.4) and 4-OH-E2 (RBA ) 52.9), 6R- ) ) of estriol (E3) and ethynylestradiol (EE). In both cases, OH-E2 (RBA 0.71), and E3 (RBA 9.72) were less - the fold reduction in activity for stronger estrogens (E2 active than E2 with 2 140-fold differences. The only and EE) was about 6 times lower than the fold reduction positions where introduction of a substituent does not R R for relatively weak estrogens (E1 and E3). This indicates interfere with binding are 7 ,11â, and 17 (6). The that the phenolic ring is more critical for weak estrogens degree of increase or decrease in activity is strongly than strong ones. dependent on the substituents in those positions. The 7R- The lack of a phenolic ring is a likely explanation for and 11â-positions are structurally equivalent for steroids why most other steroids such as androgens, progester- and can bear large substituents. Small steric substituents ones, and cholesterols were inactive in ER binding, or introduced at these positions generally increase activity exhibited very low activity, such as (29). Large substituents reduce the RBA and give rise to antiestrogenic activity, such as ICI 182,780 (RBA ) 37.5) (Figure 1B). Clearly, the precise distance (d - ) between O O ) R the 3- and 17â-OH groups of steroids, as well as their and ICI 164,384 (RBA 14.5). A 17 -ethynyl substituent orientation, governs binding affinity. For examples, is favorable for ER binding. EE is a stronger estrogen ) ) because the two OH groups had the same distance and (RBA 190) than E2 (RBA 100). Norethynodrel (RBA ) 0.22) is a progesterone derivative with a 17R-ethynyl orientation as those of E2 (dO-O ) 11.0 Å), 3â-androstane- substitution, and is a popular oral contraceptive drug. It diol (RBA ) 0.12, dO-O ) 11.3 Å) was one of the most active steroids that does not contain a phenolic ring. Also, is a moderately strong binder (454-fold weaker than E2). 17R-E2 (RBA ) 3.07, dO-O ) 10.4 Å) has a RBA 33-fold DES-like Chemicals. DES (RBA ) 400) is one of the lower than that of E2, and the RBA of 3R-androstanediol highest-affinity synthetic estrogens. Its activity is 4 times 284 Chem. Res. Toxicol., Vol. 14, No. 3, 2001 Fang et al.

negative way. The less energy a molecule sacrifices during the binding, the stronger the binding affinity will be. E2 has a much larger energy difference (∆Hf ) 40.5 kcal/mol) than DES (∆Hf ) 17.8 kcal/mol), which might help explain why DES is a better binder. The O-O distance is 1 Å larger in DES (dO-O ) 12.1 Å) than in E2 and functionally resembles the 3- and 17â- OH’s of E2. The loss of the H-bonding capability of one OH group reduces the RBA. For example, if one of the OH groups is converted to its methyl ether, DES mono- methyl ether (RBA ) 20.43), the RBA decreases ∼20- fold (Figure 2A). A similar activity reduction was also observed for hexestrol (RBA ) 300) (Figure 2B); replacing one of the OH groups with a methoxy group resulted in a 32-fold less active derivative, hexestrol monomethyl ether (RBA ) 9.37). The reduction in activity is much more significant when both OH groups of DES are methylated; the RBA decreases more than 7000-fold for DES dimethyl ether (RBA ) 0.056). This demonstrates that while the loss of one H-bond donor will reduce the binding affinity by ∼20-fold, the loss of the second H-bond donors diminishes the activity by a further 350-fold. 4-Methyl-substituted hexestrol [meso-p-(R,â-diethyl-p- methylphenethyl)phenol, RBA ) 4.00] and hexestrol monomethyl ether (RBA ) 9.37) have comparable binding affinities. This is consistent with the observation that 17- deoxy-E2 has an RBA close to that of E1. In both cases, the chemicals with or without a H-bond acceptor exhibit similar RBAs. It suggests that with a phenolic ring at one end, the contribution of the OH group at the other end is mainly through H-bond donor interaction with the receptor. The degree of contributions to the binding activity of the two ethyl groups of DES is as significant as its OH groups. DES (RBA ) 400), dimethylstilbestrol (DMS, RBA ) 14.50), and 4,4′-dihydroxystilbene (RBA ) 0.28) are in sequence one carbon atom shorter at the two ethylene side chains. The RBAs of these latter two chemicals were 28- and 1423-fold lower than that of DES. The ethyl groups may contribute to ER binding in three distinct ways. First, they increase the molecule’s hydro- phobicity. The hydrophobicity of DES (log P ) 5.64) is larger than that of E2 (log P ) 3.94). The two ethyl groups contribute more than one-third of the hydrophobicity of the overall structure. Second, they maintain the rigid binding conformation of DES. The decrease in the length of the side chains introduces more flexibility for both phenolic rings. Third, probably the most important contribution of the ethyl groups to binding is to occupy the precise space that interacts with the receptor binding Figure 2. (A) DES, (B) hexestrol, and (C) triphenylethylene pocket site, which resembles the 11â- and 7R-substituents derivatives. of E2, and therefore increase its RBA. It is evident that the attachment of small alkyl substituents to the 11â- higher than that of E2. The symmetry of the molecule, position can lead to a considerable increase in binding the distance between the two OH groups, and the two affinity for E2, such as 11â-chloromethyl-, ethyl-, and ethyl side chains provide precisely correct spacing for vinyl-E2 (28) with RBA values in the range of 120-230. hydrophobic and H-bond interactions (21). On the basis of the recent crystal structure of the E2- To efficiently bind to a receptor, the ligand tends to and DES-ER complexes, both ligands could be aligned adopt a conformation that generally is not the one with together by overlaying their receptor coordinates. By the lowest energy. The free ligand which is in a stable modifying the ethyl group to the 11â-position of the E2 conformation adjusts to a less stable conformation to binding conformation, we have been able to compare the achieve optimal binding. Nevertheless, ligand binding to relative position of the ethyl groups from 11â-ethyl-E2 the ER will release energy in general. The energy with that of DES. As shown in Figure 8, the ethyl groups difference of a molecule, between the conformation in its from both molecules are positioned extremely close to binding mode and that in the minimum conformation of each other, and fit well in the hydrophobic binding the free ligand, contributes to the binding affinity in a pocket. This observation also suggests that the specific SARs of Diverse Estrogens Chem. Res. Toxicol., Vol. 14, No. 3, 2001 285

Figure 3. (A) Flavones, (B) flavanones, (C) isoflavones, (D) coumestans, (E) chalconoids, and (F) mycoestrogens. orientation of the two ethyl groups of DES is critical to ether, and 4,4′-dihydroxystilbene versus 4,4′-ethylene- its function. That may explain why meso-hexestrol (RBA diphenol. However, 4-phenylethylphenol (RBA ) 0.002) ) 300) was 80-fold stronger than DL-hexestrol (RBA ) with single phenol group was more active than 4-stilbenol 3.60). (slight binder). These observations indicate that when In general, DES derivatives were stronger binders than both OH groups contribute to the binding, a rigid hexestrol derivatives, because the former has a rigid structure is critical for a better fit to the ER. However, double bond connecting the two benzene rings. This is when a chemical contains only one phenolic ring, binding evident from the RBA comparison of three pairs of is dependent on how well the rest of the structure fits chemicals between these two chemical classes: DES into the binding pocket and the binding activity is more versus hexestrol, DES monoether versus hexestrol mono- favorable to a chemical with certain flexibility. 286 Chem. Res. Toxicol., Vol. 14, No. 3, 2001 Fang et al.

Figure 5. (A) PCBs and (B) nonchlorinated biphenyls.

vonoids, coumestans, chalconoids, and mycoestrogens. The binding activity of these four classes follows this order: mycoestrogens > coumestans > flavonoids > chalconoids. The representative chemicals for each class are R-zearalenol (RBA ) 43), coumestrol (RBA ) 0.9), (RBA ) 0.45), and (RBA ) 0.07). Flavonoids are the largest class of phytoestrogens, which include flavones (Figure 3A), flavanones (Figure 3B), and isoflavone (Figure 3C). Their basic construction consists of a benzene ring directly connected to C2 or C3 of chromone or 4-chromanone. The positions of the A and C benzene rings of flavonoids resemble those in DES, where they are separated by two atoms in a trans conformation. Thus, the binding affinity of flavonoids is largely dependent on the relative positions of the OH groups at the A and C rings, respectively, to mimic the 4,4′-OH of DES. It is generally considered that isofla- vones are more active than flavones (30). That is because most isoflavones, such as genistein, , and , have two OH groups at the 7- and 4′-positions that match the 4- and 4′-positions of DES, respectively (Figure 3C). Although most flavones such as , , fisetin, and morin also have the 7- and 4′-positions Figure 4. (A) Diphenolalkanes, (B) benzophenones, and (C) occupied by OH groups, these two positions correspond DDTs. to the 3- and 4′-positions of DES (Figure 3A). Actually, the 6- and 4′-positions of flavones match the 4- and 4′- (Figure 2C), which act as anti- positions of DES, respectively. This was evident because estrogens, are structurally similar to DES, but have an 3,6,4′-trihydroxyflavone (RBA ) 0.45) and 6,4′-dihydoxy- additional phenyl group attached at the ethylene moiety. flavone (RBA ) 0.15) have RBAs comparable to those of The RBA value, presented here, cannot distinguish isoflavones. agonist from antagonist activity. 4-OH-tamoxifen (RBA Like the 4- and 4′-OH groups of DES, the 6- and 4′- ) 175.24) binds to the ER 100 times stronger than OH groups of flavonoids (7- and 4′-OH in isoflavones) play tamoxifen (RBA ) 1.62) and 10 times stronger than a similar role in binding. Although a small difference (RBA ) 15.24) because its phenolic ring (∼2.5-fold) in RBA values was observed for the biochanin resembles the A ring of E2. If its OH group is eliminated A (RBA ) 0.0043) and (RBA ) 0.0018) pair and (tamoxifen, RBA ) 1.62) or its position is changed the 6-hydroxyflavanone (RBA ) 0.0009) and 4′-hydroxy- (droloxifene, RBA ) 15.24), the binding affinity is re- flavanone (RBA ) 0.0023) pair, it is difficult to eliminate duced. For clomiphene, a chloro substituent at the the possibility that both phenolic rings (A and C) of ethylene and, for , a chloro substituent at the flavonoids can functionally mimic the phenolic A ring of ethyl side chain both produce binding affinities similar E2. This suggests the possibility that flavonoids may bind to that of tamoxifen. to the ER in either of two orientations that differ by 180°. Phytoestrogens. There are four major structurally The recent description of the crystal structure for the distinct chemical classes of phytoestrogens (10): fla- genistein-ERâ complex revealed that its phenolic C ring SARs of Diverse Estrogens Chem. Res. Toxicol., Vol. 14, No. 3, 2001 287

Figure 7. Miscellaneous chemicals.

Figure 8. Superimposed structures of 11â-ethyl-E2 (green) and DES (yellow). E2 and DES are first aligned on the basis of the superposition of the receptor coordinates of the crystal struc- - - Figure 6. (A) Alkylphenols, (B) parabens, and (C) alkyloxy- tures between the E2 and DES ER complexes. The 11â- phenols. position of the E2 is then modified by adding an ethyl group to form 11â-. The ethyl groups of 11â-ethyl-E2 and DES are positioned closely in a hydrophobic binding pocket. served the same function as the A ring of E2 to form the H-bond interaction with the receptor (31). Like DES derivatives that have larger RBAs than Converting one of the OH groups to a methyl ether hexestrol derivatives, flavones are slightly better binders significantly reduced the RBA of isoflavones. Prunetin ) ) than flavanones because their benzene rings are attached (RBA 0.0018) and (RBA 0.0043) have to the double bond of the B ring of chromone, which leads 250- and 100-fold lower RBAs, respectively, than their ) to a more rigid and flat structure. Within the same parent compound, genistein (RBA 0.45), while the RBA ) structural frame, the flavone apigenin (RBA ) 0.028) has of (RBA 0.0013) is 17-fold lower than that an RBA 4 times higher than that of the flavanone of daizein. (RBA ) 0.0075). Similarly, genistein was Unlike the fact that 1-OH-E2 is less potent than E2 (33), more potent than 2,3-dihydroxygenistein (32). However, an additional OH group in position 5 of isoflavones, which the result is only applicable for flavonoids containing both resembles the 1-position of E2, increases estrogenic OH groups at the A and C rings. The relative activity is activity (10, 34). Specifically, genistein is 20-fold more reversed between 6-hydroxyflavanone (RBA ) 0.0009) active than daizein, and biochanin A is 4-fold more active and 6-hydroxyflavone (RBA ) 0.0004), which are consis- than formononetin. This may be explained by the forma- tent with observations on DES-like chemicals. tion of an intramolecular H-bond between the 5-OH and Generally, a large steric hindrance will reduce activity. the carbonyl groups which enhances the electron with- Genistein and naringin are inactive compared to genistein drawal of the carbonyl group and lead to a better 7-OH (RBA ) 0.45) and naringenin (RBA ) 0.0075). However, H-bond donor. compared to (NA), rutin exhibited a very low Two coumestans were assayed, and both had RBAs activity. This indicates that the 3-position of flavones is ∼100-fold lower than that of E2 (Figure 3D). While equivalent to the 7R-position of E2 where a large group coumestrol has a relatively rigid and flat structure that can be fitted in the binding pocket. is similar to E2, 4-ethyl-7-hydroxy-3-(p-methoxyphenyl)- 288 Chem. Res. Toxicol., Vol. 14, No. 3, 2001 Fang et al.

Table 3. Relationship between the O-O Distance and Table 4. Hydrophobicity (log P) of DDTs the RBA DDT log P DDT log P DDT log P RBA d - (Å) mycoestrogen RBA d - (Å) O O O O o,p′-DDT 6.79 o,p′-DDD 5.87 o,p′-DDE 6.00 R-zearalanol 30 11.00 â-zearalenol 0.20 9.85 p,p′-DDT 6.91 p,p′-DDD 6.02 p,p′-DDE 6.51 R-zearalenol 43 11.28 2.1 9.77 â-zearalanol 0.64 9.23 olefin (RBA ) 2.47) binds 10 times stronger than mono- hydroxymethoxychlor olefin (RBA ) 0.23). This suggests 2H-1-benzopyran-2-one has an ethyl group that is func- that the second phenolic ring of olefin tionally similar to that of DES. The latter had the same derivatives contributes to the binding, which may be RBA as coumestrol because its ethyl group favors binding analogous to the 17â-OH of E2. The crystal structures of which compensates for the reduced RBA form converting four ligands with the ER reveal that a considerable one of the OH groups to a methyl ether. flexibility of the His 524 residue of the receptor permits The generic structure of chalconoids is two benzene H-bond interactions with ligands in a wide range of - rings connected by three carbons (Figure 3E). The O O relative positions. Because of the sp2 hybridization, the distance between the two OH groups in the B ring and angle between the two phenyls in olefins (120° with sp2) ∼ ∼ the one in the A ring is 11 and 12 Å, respectively, are larger than that in the non-olefin methoxychlors (109° which is comparable to those of DES and E2. Since their with sp3). The distance between the two O atoms for flexible molecular structures do not favor binding, chal- olefin derivatives is ∼9.7 Å, and that for the non-olefins conoids have RBAs that are 1000-fold lower than that of is ∼9.3 Å. Therefore, the O-O distance for olefin methoxy- E2. The elimination of one of the OH groups on one side, chlors is closer to that in E and more favorable for ′ 2 such as 4-hydroxychalcone and 4 -hydroxychalcone, caused binding. a loss in activity of ∼20-fold. The results further demon- strate that two OH groups within a precise distance play For diphenolalkanes, changes in chain length at the an important role in ER binding. bridging carbon lead to the RBA changes: bisphenol B (RBA ) 0.086) > (RBA ) 0.008) > bisphenol Mycoestrogens are the most active chemicals in phyto- ) estrogens (Figure 3F). As shown in Table 3, R-zearalenol F (RBA 0.0009). The longer the side chain, the greater and R-zearalanol had RBAs 2 orders of magnitude higher the binding affinity for the ER (35). Common pharmaco- than those of their â-isomers. The activity of zearalanone phore identification indicated that the optimal structural was between those of its R- and â-isomers. The order of superposition between bisphenol A and E2 is one in which their RBAs was consistent with that of their O-O the two benzene rings of bisphenol A are positioned over distances. In R-zearalenol and R-zearalanol, these were the A and C rings of E2, with the bridge carbon overlaying about 11.0 Å, the same distance found between the 3- the B ring (Figure 9). Thus, the effect of steric substit- uents on the bridge atom of diphenylmethanes on ER and 17â-OH groups of E2; â-zearalanol and â-zeralenol R are 1 Å shorter (Table 3). This indicates that their RBAs binding is analogous to the effect of 7 -substituents on are dependent on one of the critical structural features, E2 binding. the O-O distance, when chemicals have the same DDT, DDD, and DDE isomers have structural frame- structural frameworks. works similar to that of bisphenol A (Figure 4C). The o,p′- Diphenylmethanes. The generic structure for this isomers are active in binding, while p,p′-isomers are not. chemical class is two benzene rings separated by one The orthochlorine of o,p′-isomers mimics the steric 11â- carbon (or other atoms) (Figure 4). The 4-OH substituent substituent of E2 and increases structural rigidity, which is critical for binding, which is supported by a number favors binding. DDT isomers have greater hydrophobicity of observations: (log P) than either DDD or DDE isomers (Table 4), which All chemicals with a 4-OH substituent exhibit binding might explain why o,p′-DDT is the strongest binder activity, except 4,4′-methylenebis(2,6-di-tert-butylphenol) among the six DDT derivatives that were evaluated. that contains four steric hindered tert-butyl groups ortho Biphenyls. 4-OH-PCBs tend to be good binders to the 4- and 4′-OH groups. These prevent H-bond (Figure 5A), which is consistent with observations in interactions between the phenolic rings and the receptor. other chemical classes. As the number of chloro substitu- Two benzophenones (Figure 4B), 2,4-dihydroxy- tions at the nonphenolic ring increases, more electron benzophenone and 4,4′-dihydroxybenzophenone, both withdrawal is found in the phenolic ring, which results containing a 4-OH group, exhibit weak binding activity. in higher pKa values (36) and a better H-bond donor. However, 2,2′-dihydroxybenzophenone is inactive due to 2′,3′,4′,5′-Tetrachloro-4-biphenylol and 2′,5′-dichloro-4- the absence of a 4-OH group. Similarly, 2,2′-dihydroxy- biphenylol are the strongest binders in the group. Both 4-methoxybenzophenone and 2-OH-4-methoxybenzo- chemicals have an orthochlorine substitution at the phenone are inactive because of the methylation of the nonphenolic ring. Korach et al. (9) reported that PCB 4-OH. compounds with the strongest affinities possess either Among the 12 DDTs (Figure 4C), the four with 4-OH single- or multi-orthochlorine substitution. The ortho substitution show much stronger binding than the others. substitution restricts the conformational flexibility of Methoxychlor and methoxychlor olefin are inactive be- PCBs, which favors binding. However, our data show cause of methylation of their OH groups. that a considerable improvement of binding is only Bisphenol A (RBA ) 0.008) and cumyl phenol (RBA ) associated with the orthochlorine substitution at the 0.005) have comparable binding affinities, as do mono- nonphenolic ring. For example, 2-chloro-4-biphenylol hydroxymethoxychlor (RBA ) 0.13) and HPTE (RBA ) (RBA ) 0.002) has binding affinity similar to that of 4′- 0.25). It suggests that only one phenolic ring is critical chloro-4-biphenylol (RBA ) 0.007), but 28 times lower in binding for chemicals with the sp3 hybrid bridged than that of 2′,5′-dichloro-4-biphenylol (RBA ) 0.036). carbon. However, this is not the case for chemicals with Analyzing common pharmacophores between E2 and the sp2 hybrid bridged carbon. Dihydroxymethoxylchlor 2′,3′,4′,5′-tetrachloro-4-biphenylol (Figure 9), we found SARs of Diverse Estrogens Chem. Res. Toxicol., Vol. 14, No. 3, 2001 289

Figure 10. Correlation of log RBA with log P for phenols.

also with possibly acting as a hydrophobic group resem- bling the B ring and/or 11â-substituent of E2. The phenylphenols (Figure 5B) are relatively lower affinity binders than the 4-OH-PCBs. With the change of OH position from 4 to 3 and to 2, the RBA decreases in the following order due to the decrease in the accessi- bility of the OH group for H-bond interaction: 4-phenyl- phenol > 3-phenylphenol > 2-phenylphenol. Phenols. Alkylphenols (Figure 6A), parabens (Figure 6B), and alkyloxyphenols (Figure 6C) are contained in this class; nonylphenol has the highest RBA (0.031). The activities of the phenols are largely dependent on the alkyl chain length at the para position. 2-Ethylhexyl- (RBA ) 0.018) has an RBA close to that of 4-tert-octylphenol (RBA ) 0.015), and a chain length with the same number of carbons (eight). The activity of parabens follows this order: 2-ethylhexyl > heptyl > benzyl > butyl > propyl ) ethyl > methyl. A similar trend is also observed in alkylphenols (11, 23). The relationship of ER binding activity (log RBA) with hydrophobicity (log P) is linearly correlated for this class (r2 ) 0.82, log RBA ) 0.58 log P - 5.20; Figure 10). Each unit of log P change will cause 0.58 unit of logRBA change, demonstrating the importance of hydrophobicity for binding. However, dodecylphenol is an outlier, indi- cating that this linear correlation is only valid within a certain range. Miscellaneous. A number of miscellaneous chemicals

Figure 9. Four-point pharmacophore of E2, the 3-OH and the in the NCTR data set were identified to be active in ER centers of rings A, B, and C, identified three structurally diverse binding (Figure 7). Some have similar structural frame- xenoestrogens, bisphenol A, 2′,3′,4′,5′-tetrachloro-4-biphenylol, works as previously discussed structural classes, while and ketone. others do not. Doisynolic and allenolic acids are nonsteroidal estro- that the two benzene rings of PCBs match the A and C gens. Their binding affinities for the ER are normally ring of E2 very well while the orthochlorine occupies the 100-fold lower than that of E2, but their in vivo potencies B ring position. For PCBs having two orthochlorine are high because of a long duration of action (37). Two of substitutions at the nonphenolic ring, the second ortho- their derivatives, R,R-dimethyl-â-ethylallenolic acid and chlorine is most likely to overlay the 11â-position of E2 doisynoestrol, were tested; both exhibited an ability to where the introduction of small steric substituents compete E2 for ER binding. Their OH and acid groups normally improves binding. Thus, these types of PCBs as well as their naphthalene substructures match the 3- are better binders. It is evident that 2′,4′,6′-trichloro-4- and 17â-OH and A and B rings of E2, respectively, very biphenylol is more active than 2′,3′,4′,5′-tetrachloro-4- well in 3D space.3 Because of the methylation of its OH biphenylol (9). It is safe to conclude that the contributions of orthochlorine substitution to binding are associated not 3 X. Qian, H. Fang, H. Hong, W. Tong, R. Perkins, and D. M. only with the restriction of conformational flexibility but Sheehan, unpublished results. 290 Chem. Res. Toxicol., Vol. 14, No. 3, 2001 Fang et al. group, doisynoestrol has activity ∼100-fold lower than neuromuscular system, liver, etc. (45). Kepone (RBA ) that of doisynolic acid, which is consistent with the 0.013) was found to be 10 000 times weaker than E2 in observations for steroids. its affinity for the ER. How its structural features relate 4,4′-Methylenedianiline and 4-aminophenyl ether struc- to E2 in binding is still puzzling. Kepone contains four turally resemble , where the OH groups are cyclopentanes in the left, right, top, and bottom sides replaced with the amino groups. Both chemicals are fused together to form a cagelike structure (Figure 7). It inactive in binding. A similar result is also observed for appears that the carbonyl group at the C2-position is butyl-2-aminobenzoate, which is structurally similar to essential because mirex, its analogue that has the car- butylparaben. This evidence reinforces the importance bonyl group replaced with two chlorine atoms, is inactive of the phenolic ring in binding, particularly its essential in binding. Common pharmacophore identification indi- role for weak estrogens. cates that kepone is similar to E2 in various structural Benzil is a selective inducer and a potent in vitro features: the carbonyl group along with the hydrophobic activator of microsomal epoxide hydrolase (38). It is also center of the 2,3,3a,5b,1a-cyclopentane ring, C4-Cl, and used as a photoinitiator for the visible light polymerizing the hydrophobic center of the two Cl atoms at the C5- resin system that is widely applied in modern dentistry position mimics the function of the phenolic A, B, and C (39). Benzil has a structural resemblance to stilbene and rings in E2 (Figure 9). It is important to point out that has been studied as a new type of inducer of drug- kepone can rapidly pick up moisture to form kepone metabolizing enzyme along with stilbene oxide (40). The hydrate that should be more active for the ER (46). 4,4′-hydroxylated benzils share a similar structural construction with the 4,4′-hydroxylated stilbenes: two Discussion OH groups separated by two benzene rings with two sp2 (carbonyl) carbons. Their close structural similarity In principle, the biological activity of a chemical is leads to similar binding affinities found for 2,2′,4,4′- determined by its structure. The chemical structure can tetrahydroxybenzil (RBA ) 0.209) (Figure 7) and 4,4′- be represented in three different general ways: 2D dihydroxystilbene (RBA ) 0.281) (Figure 2A). substructures, 3D pharmacophores, and physicochemical Nordihydroguariaretic acid is an antioxidant for fats properties. A 2D substructure is a structural fragment and oils in foods. It is a flexible molecule with two of a molecule, which often can be used as a strong catechols separated by four carbons. According to our indicator of a particular activity, such as the phenolic ring recent study on common pharmacophore identification for the ER. A 3D pharmacophore is a portion of a of estrogens,3 nordihydroguariatic acid shares structural chemical’s 3D structure that is considered essential in commonalities with E2: its two catechol groups mimic eliciting the biological activity of interest, such as the - E2’s A and D rings, while its bridged carbons fold in such precise O O distance, the orientation of the OH group, a way to occupy the space recognized by 7R-substituents and the location of the hydrophobic center. A physico- of E2. Normally, hydroxylation of phenol to create cate- chemical property of a molecule is a measure of one chols reduces estrogenic activity (34), and the flexible property of a whole molecule. For example, log P mea- structure is not favorable for binding. Therefore, even sures a chemical’s hydrophobicity. ER binding activity though nordihydroguariatic acid shares the key phar- relates to all these structural features. The SAR studies macophores of E2, its RBA is much lower. for ER binding by individual chemical classes indicate Siloxanes are widely used for industrial and consumer that some features may well represent binding depend- product applications (41). A single subcutaneous injection encies for one structural class, while other features may in mice using samples from breast implant or poly- better represent binding dependencies for a different (dimethylsiloxane) results in the wide distribution of the structural class. These structural features are inherently low-molecular weight siloxanes throughout the body (42). related, suggesting that structural commonality exists Studies by Bennett et al. (43) have shown that a series among structurally diverse estrogens. of low-molecular weight siloxanes alter male reproductive It is well-accepted that the precise spacing of two OH function in a number of mammalian species. An estro- groups at either end of an essential planar and primarily genic effect on the immature rat uterus has also been hydrophobic molecule is considered the structural basis demonstrated for several organosiloxane compounds for ER binding (47). The E2-ER H-bonding network (44). Two linear disiloxanes were included for this (Figure 11) demonstrates the critical role of H-bonding study. Neither chemical possesses H-bonding capability ability for a chemical to bind to the ER (20). Although at either end of the structure which is normally con- the “anchor-like” H-bonding network at the A ring of E2 sidered to be essential for a chemical to bind to the imposes an absolute requirement that an effective ligand ER. However, 1,3-diphenyltetramethyldisiloxane exhibits must contain a phenolic ring, the remainder of the marginal activity. Chalcone (RBA ) 0.0015), 1,3-diphenyl- binding pocket can accept a number of different hydro- tetramethyldisiloxane (RBA ) 0.0007), and butylbenzyl- phobic groups. By comparing various estrogen classes phthalate (slight binder) share similar structural frame- through SAR studies, we can summarize the general works: two benzene rings (hydrophobic centers) sepa- structural requirements relevant to the template E2 rated by three atoms. Their activities are proportional structure: (1) H-bonding ability of the phenolic ring to the rigidity of their backbone, suggesting that the mimicking the 3-OH, (2) H-bond donor mimicking the precise distance of two hydrophobic centers with a rigid 17â-OH and O-O distance between 3- and 17â-OH, (3) linkage favors binding. precise steric hydrophobic centers mimicking steric 7R- Of the variety of estrogens whose structures have little and 11â-substituents, (4) hydrophobicity, and (5) a ring resemblance to E2, kepone is of particular interest for structure. These important features provide the struc- many reasons. Not only does it produce a variety of tural basis for a xenoestrogen to exhibit binding activity. “estrogen-like” effects on the female reproductive system, H-Bonding Ability of the Phenolic Ring Mimick- but it also causes other toxic effects, including in the ing the 3-OH. The importance of H-bonding has been SARs of Diverse Estrogens Chem. Res. Toxicol., Vol. 14, No. 3, 2001 291

phenol > 2-methylphenol ) 2-tert-butylphenol > 2,6- dimethylphenol > 2,6-di-tert-butylphenol (48), in which 2,6-di-tert-butylphenol is not a H-bond donor. This is consistent with the lack of binding activity observed for 4,4′-methylenebis(2,6-di-tert-butylphenol) (Figure 4A). In conclusion, a good H-bond donor mimicking the 3-OH of E2 may be more important than a H-bond acceptor for xenoestrogen. H-Bond Donor Mimicking the 17â-OH and O-O Distance between 3- and 17â-OH. Chemicals could be strong estrogens if they possess a phenolic ring and an additional OH group within a certain distance range. Chemicals containing only one phenolic group are most likely to be weak to medium ER ligands. The 17â-OH of E2 makes a single H-bond as a donor with His 524, because the two N atoms of imidazole act as an H-bond acceptor in most cases. This is consistent with the Figure 11. ER-E2 hydrogen-bonding network (52). observation of similar RBAs for the E1 and 17-deoxy-E2 Table 5. H-Bond Distances between Two OH Groups of (Figure 1A) pair and the meso-p-(R,â-diethyl-p-methyl- Four Ligands and the Receptor phenethyl)phenol and hexestrol monomethyl ether pair 3-OH 17-OH (Figure 2B). Therefore, a stronger H-bond donor, such Glu 353 Arg 394 water His 524 as the phenolic OH, favors binding at the position that ER ligand [dO-O (Å)] [dO-N (Å)] [dO-O (Å)] [dO-N (Å)] resembles the 17â-OH of E2. The imidazole of His 524 is not only able to rotate E2 2.37 3.08 2.82 2.77 DES 2.51 3.33 3.19 2.67 (because of the alkyl chain), but its two equivalent raloxifene 2.38 2.93 2.97 2.62 imidazole N atoms could compensate for the change in 4-OH-tamoxifen 2.45 3.04 3.07 the oxygen position of 17â-OH and maintain a favorable H-binding position. This was recognized in the E2-and recognized in most biological systems. The O atom is one raloxifene-ER crystal structure. The phenolic A rings of of the most important heteroatoms directly associated both molecules overlap quite well, whereas the 17â-OH with H-bonding. Of the 130 active chemicals in the NCTR group of E2 and the OH group of raloxifene are 5 Å apart. data set, 125 chemicals contain an O atom and 3 However, these two OH groups have a similar distance chemicals contain a Cl atom that may serve as a weak to the N atom of the imidazole (Table 5). It suggests that H-bond acceptor. Even though the degree of specificity the flexible His 524 is more tolerant of the second OH is reduced because of two active chemicals without groups of a chemical in forming an effective H-bond H-bonding ability that produce specific steric effects, it interaction. Although the H-bonding strength is depend- is difficult to avoid the conclusion that the requirement ent on the relative positions (angle and distance) between of H-bonding ability is the most important characteristic the imidazole of His 524 and the OH group, the O-O for estrogen. distance could provide a fair estimation of the H-bonding It has been long understood that the phenolic ring ability of the second OH group analogous to the 17â-OH is normally associated with estrogenic activity. One of E2. hundred eight of 130 active chemicals (83%) in the For purposes of analysis, we have divided the NCTR NCTR data set contain a phenolic ring. The contribution data set into four activity categories: strong (RBA > 1), of the phenolic ring in binding is much more significant medium (1 > RBA > 0.01), weak (0.01 > RBA > than any other structural feature. The current crystal detectable activity), and inactive. As shown in Figure 12, structures of four ligand-ER complexes reveal that the most strong to medium ER ligands contain two OH phenolic rings of all four ligands are closely positioned groups with an O-O distance ranging from 9.7 to 12.3 at the same location to allow H-bond interactions with Å. The chemicals with an O-O distance not in this range Glu 353, Arg 394 of the receptor, and a water molecule. are highly likely to be weak estrogens. The distance of the phenolic O atom of these four ligands Precise Steric Hydrophobic Centers Mimicking to Glu 353, Arg 394, and the water molecule is sum- Steric 7r- and 11â-Substituents. The volume of marized in Table 5. It appears that the H-bonding of the the ER binding pocket (450 Å3) is about twice that of 3 phenolic OH is much stronger with Glu 353 than Arg 394 E2 (245 Å )(20). The length and breadth of the E2 and the water molecule because of the shorter distance. skeleton are well matched by the receptor, but there are Thus, the phenolic OH contributes to binding largely large unoccupied cavities at the 7R- and 11â-positions of through its H-bond donor ability because of the H-bond E2. The positions of these cavities allow steric groups of acceptor properties of Glu 353. The phenolic OH is a certain sizes to fit, and are of great importance for better H-bond donor than an acceptor. Replacing OH with xenoestrogens, including DES-like chemicals, diphenyl- the weak H-bond donor NH2 reduces activity for E2 (17), methanes, and biphenyls. For example, the binding and diminishes activity for bisphenol F and butylparaben affinities of DES, DMS, and 4,4′-dihydroxystilbene change (4-aminophenyl ether and butyl-4-aminobenzoate, Figure dramatically with changes in the side chain length on 7). It is important to point out that the H-bond donor the C-C double bond. This trend was also observed (4) ability of the phenolic OH is dependent on a number of in in vivo activity; 4,4′-dihydroxystilbene had 250-fold factors, particularly the nature of ortho substituents that lower potency than DMS and 16700-fold lower potency affect the OH accessibility. The H-bond donor ability for than DES. Even triphenylethylene (RBA ) 0.002) (Figure several ortho-substituented phenols follows this trend: 2C) that has no H-bonding ability still exhibits weak 292 Chem. Res. Toxicol., Vol. 14, No. 3, 2001 Fang et al.

suggests that log P is only important in binding when the key pharmacophores are in place. This can be further demonstrated by plotting mean log P values versus activity categories for chemicals containing two OH groups with distances between 9.7 and 12.3 Å (mimicking the 3- and 17â-OH of E2) and only one 3-OH group. As shown in Figure 13, the positive effect of hydrophobicity on binding is much more apparent for chemicals contain- ing the 3- and 17â-OH, and a 3-OH. Interestingly, for each activity category, the mean log P value of chemicals with one OH group is larger by almost the same amount compared to that of chemicals containing two OH groups mimicking the 3- and 17â-OH of E2. This indicates that chemicals lacking the effective O-O distance require greater hydrophobicity to reach binding activities similar - Figure 12. Effect of the range of the O O distance on RBA. to those of chemicals with the 3- and 17â-OH groups. The activity categories were arbitrarily defined as follows. For strong estrogens, RBA > 1. For medium estrogens, the RBA is Ring Structure. Effective ER ligands must possess a between 0.01 and 1. For the weak estrogens, RBA is between ring structure. A literature survey of more than 2000 0.01 and a detectable activity. Inactive chemicals show no unique chemicals that were tested by various in vitro detectable activity in the assay. The RBA for E2 is set to 100. and/or in vivo assays for estrogenic activity reveals that no estrogenic chemical has been found that does not possess a ring structure. The ring construction increases the rigidity of both the structure and the steric center, which favors ER binding. It appears that the flat aro- matic ring, which fits better to the narrow “tunnel-like” arrangement in the receptor for the phenolic A ring, is more important than other ring structures. The majority of xenoestrogens contain at least one aromatic ring. Only five chemicals without aromatic rings were found to be active, of which four are steroids (Figure 1B) in addition to kepone. All five chemicals possess H-bond capability Figure 13. Relationship of the mean log P values with binding with a rigid hydrophobic backbone that matches the A, activity. The activity categories are defined in the legend of B, and C rings of E2 very well. It suggests that nonaro- Figure 12. matic estrogens require several key structural features to exhibit binding activity, at least including a strong binding. Clearly, the precise location of steric bulk in the electronegative atom (O, S, N, etc.) for H-bond interaction region of the 7R- and/or 11â-positions of E2 is critical for and a rigid hydrophobic backbone. binding. It is evident that DL-hexestrol’s RBA activity is There are several substructures inherently associated about 100-fold lower than that of mesohexestrol. The two with ER binding activity. The common structural frame- isomers only differ in the direction of the two ethyl work for most estrogens contains an aromatic ring, substitutes. Similarly, indenestrol A was found to have normally a phenolic ring, separated from a hydrophobic higher affinity for the ER (49) than the enantiomers of center that mimics one of the centers of E2’s B, C, or D indenestrol B (50). ring. PCBs have two benzene rings connected directly, Hydrophobicity. The ER ligand-binding pocket has which resemble the A and C rings. The ortho substituents a 3D “cross”-like shape; the center and vertical ends are of PCBs match the B ring and/or 11â-substituents of E2 mainly hydrophobic, whereas the polar functions are and increase binding affinity. Bisphenol A and DDT-like located at opposite ends of the horizontal cavity. An effect chemicals have two benzene rings separated by one atom, of hydrophobicity in binding was observed for phthalates whereas two rings mimic the A and C ring and the (51). Hydrophobicity (log P), expressed as a ratio of substituents on the bridge atom match the B ring. The solubility between octanol and water, is important for two benzene rings of DES derivatives resemble the A and many biological end points. Log P values too high or too D rings. The binding of DES derivatives is mainly low can be associated with poor transport characteristics. governed by the presence or absence of two OH groups Unlike the aforementioned three estrogen-related struc- at the 4,4′-position and two ethyl groups to mimic the tural features that are critical in binding only by inter- 7R- and 11â-substituents of E2. It seems that two benzene acting with specific amino acids of the receptor, hydro- rings separated by three atoms provide an optimal steric phobicity is a physicochemical property. interaction with the receptor, which is demonstrated by Although the mean log P values for strong, medium, three weak estrogens, chalcone, 1,3-diphenyltetramethyl- and weak estrogens exhibited a positive trend (Figure disiloxane, and butylbenzylphthalate, which contain no 13), the inactive chemicals also have a wide range of H-bonding capability mimicking either the 3- or 17â-OH log P values. This suggests that better binders tend to of E2. have larger log P values but not vice versa. We plotted Summary. The SAR studies on 230 diverse chemicals log RBA versus log P for individual chemical classes; the reveal five structural features that are most important only linear correlation was for phenols (Figure 10). for chemical binding to ER. The representative chemical Phenols consist of only a phenolic ring serving as a for each structural category listed in Table 1 demon- H-bond anchor and an alkyl chain for hydrophobicity. The strates that the more key structural features a chemical linear correlation between log RBA and log P for phenols contains, the more active it is. These findings can be SARs of Diverse Estrogens Chem. Res. Toxicol., Vol. 14, No. 3, 2001 293

used, the primary step is to identify ER-related structural characteristics. The knowledge then could be used to design alignment for 3D QSAR, to select ER-related structural descriptors for QSAR, and to identify key structural features for pharmacophore models. All this tells us that structure-activity relationship analyses of xenoestrogens are essential.

Acknowledgment. We gratefully acknowledge the U.S. Department of Energy and the U.S. Food and Drug Administration (FDA) for postdoctoral support through the Oak Ridge Institute for Science and Education. This work was partially supported by the FDA’s Office of Women’s Health and the Chemical Manufacturer’s Association.

References

(1) Kavlock, R. J., Daston, G. P., DeRosa, C., Fenner-Crisp, P., Gray, L. E., Kaattari, S., Lucier, G., Luster, M., Mac, M. J., Maczka, C., Miller, R., Moore, J., Rolland, R., Scott, G., Sheehan, D. M., Sinks, T., and Tilson, H. A. (1996) Research needed for the risk assessment of health and environmental effects of endocrine disruptors: a report of the U.S. EPA-sponsored workshop. Envi- ron. Health Perspect. 104 (Suppl. 4), 715-740. (2) Patlak, M. (1996) A testing deadline for endocrine disrupters. Figure 14. Flowchart for identification of ER ligands. Environ. Sci. Technol. 30, 540A-544A. (3) Katzenellenbogen, J. A. (1995) The structural pervasiveness generalized as a set of “if-then” rules for guidance in of estrogenic activity. Environ. Health Perspect. 103 (Suppl. 7), - identifying potential ER ligands, which is depicted in 99 101. (4) Dodds, E., Goldberg, L., and Lawson, W. (1938) Oestrogenic Figure 14: activity of alkylated stilboestrols. Nature 142, 34. (1) If a chemical contains no ring structure, then it is (5) Dodds, E., and Lawson, W. (1938) Molecular structure in relation unlikely to be an ER ligand. to oestrogenic activity. Compounds without a phenanthrene nucleus. Proc. R. Soc. 125, 222-232. (2) If a chemical has a nonaromatic ring structure, then (6) Anstead, G. M., Carlson, K. E., and Katzenellenbogen, J. A. (1997) it is unlikely to be an ER ligand if it does not contain an The estradiol pharmacophore: ligand structure-estrogen receptor O, S, N, or other heteroatom for H-bonding. Otherwise, binding affinity relationships and a model for the receptor binding site. Steroids 62, 268-303. its binding potential is dependent on the existence of the (7) Dodge, J., and Jones, C. (1999) Non-steroidal estrogens. In key structural features. Kepone, dihydrotestosterone, Estrogens and antiestrogens I: Physiology and Mechanisms of norethynodrel, and 3R- and 3â-androstanediol are active Avtion of estrogens and antiestrogens (Oettel, M., and Schillinger, ER ligands that fall into this category. E., Eds.) pp 43-53, Springer-Verlag, Berlin. (8) Sadler, B. R., Cho, S. J., Ishaq, K. S., Chae, K., and Korach, K. S. (3) If a chemical has a non-OH aromatic structure, then (1998) Three-dimensional quantitative structure-activity rela- its binding potential is dependent on the existence of the tionship study of receptor ligands using the key structural features. A total of 16 chemicals in the comparative molecular field analysis/cross-validated r2-guided - NCTR data set, including o,p′-DDT, 1,3-diphenyltetra- region selection approach. J. Med. Chem. 41, 2261 2267. (9) Korach, K., Sarver, P., Chae, K., McLachlan, J., and McKinney, methyldisiloxane, 3-deoxyl-E2, mestranol, and others, fall J. (1988) Estrogen receptor-binding activity of polychlorinated into this category. hydroxybiphenyls: conformationally restricted structural probes. - (4) If a chemical contains a phenolic ring, then it tends Mol. Pharmacol. 33, 120 126. (10) Miksicek, R. J. (1994) Interaction of naturally occurring non- to be an ER ligand if it contains any additional key steroidal estrogens with expressed recombinant human estrogen structural features. For the chemicals containing a receptor. J. Biochem. Mol. Biol. 49, 153-160. phenolic ring separated from another benzene ring with (11) Routledge, E. J., and Sumpter, J. P. (1997) Structural features the number of bridge atoms ranging from none to three, of alkylphenolic chemicals associated with estrogenic activity. J. Biol. Chem. 272, 3280-3288. it will most likely be an ER ligand. (12) Grese, T. A., Cho, S., Finley, D. R., Godfrey, A. G., Jones, C. D., Lugar, C. W., III, Martin, M. J., Matsumoto, K., Pennington, L. Conclusion D., Winter, M. A., Adrian, M. D., Cole, H. W., Magee, D. E., Phillips, D. L., Rowley, E. R., Short, L. L., Glasebrook, A. L., and The goal of the study presented here was to explore Bryant, H. U. (1997) Structure-activity relationships of selective the change in ER binding activity with the change of estrogen receptor modulators: modifications to the 2-arylbenzo- thiophene core of raloxifene. J. Med. Chem. 40, 146-167. structure based on the NCTR data set, which to the (13) Tong, W., Perkins, R., Strelitz, R., Collantes, E. R., Keenan, S., best of our knowledge is so far the largest published Welsh, W. J., Branham, W. S., and Sheehan, D. M. (1997) structurally diverse data set with a consistent quantita- Quantitative structure-activity relationships (QSARs) for estro- tive binding activity measurement. The data set conveys gen binding to the estrogen receptor: predictions across species. - information far beyond the scope of this paper. The Environ. Health Perspect. 105, 1116 1124. (14) Waller, C. L., Oprea, T. I., Chae, K., Park, H. K., Korach, K. S., knowledge derived from this study enhances our under- Laws, S. C., Wiese, T. E., Kelce, W. R., and Gray, L. E., Jr. (1996) standing of the important ER binding-related structural Ligand-based identification of environmental estrogens. Chem. features, which in turn should be important in guiding Res. Toxicol. 9, 1240-1248. the development of predictive toxicology models for rapid (15) Tong, W., Perkins, R., Xing, L., Welsh, W. J., and Sheehan, D. M. (1997) QSAR models for binding of estrogenic compounds identification of xenoestrogens using computational ap- to estrogen receptor R and â subtypes. Endocrinology 138, proaches. No matter what computational approaches are 4022-4025. 294 Chem. Res. Toxicol., Vol. 14, No. 3, 2001 Fang et al.

(16) Tong, W., Lowis, D. R., Perkins, R., Chen, Y., Welsh, W. J., (35) Perez, P., Pulgar, R., Olea-Serrano, F., Villalobos, M., Rivas, A., Goddette, D. W., Heritage, T. W., and Sheehan, D. M. (1998) Metzler, M., Pedraza, V., and Olea, N. (1998) The estrogenicity Evaluation of quantitative structure-activity relationship meth- of bisphenol A-related diphenylalkanes with various substituents ods for large-scale prediction of chemicals binding to the estrogen at the central carbon and the hydroxy groups. Environ. Health receptor. J. Chem. Inf. Comput. Sci. 38, 669-677. Perspect. 106, 167-174. (17) Wiese, T. E., Polin, L. A., Palomino, E., and Brooks, S. C. (1997) (36) Ebner, K., and Braselton, W. (1985) Structural and chemical Induction of the estrogen specific mitogenic response of MCF-7 requirements of hydroxylated polychlorinated biphenyls (PCBOH) cells by selected analogues of estradiol-17 beta: a 3D QSAR study. for inhibition of energy-dependent swelling of rat liver mitochon- J. Med. Chem. 40, 3659-3669. dria. Toxicologist 5, 260. (18) Gantchev, T. G., Ali, H., and van Lier, J. E. (1994) Quantitative (37) Scribner, A. W., Jonson, S. D., Welch, M. J., and Katzenellen- structure-activity relationships/comparative molecular field analy- bogen, J. A. (1997) Synthesis, estrogen receptor binding, and sis (QSAR/CoMFA) for receptor-binding properties of halogenated tissue distribution of [18F]fluorodoisynolic acids. Nucl. Med. Biol. estradiol derivatives. J. Med. Chem. 37, 4164-4176. 24, 209-224. (19) Waller, C. L., Minor, D. L., and McKinney, J. D. (1995) Using (38) Seidegard, J., and DePierre, J. W. (1981) Benzil, a selective three-dimensional quantitative structure-activity relationships inducer and a potent in vitro activator of microsomal epoxide to examine estrogen receptor binding affinities of polychlorinated hydrolase. Adv. Exp. Med. Biol. 136, 717-728. hydroxybiphenyls. Environ. Health Perspect. 103, 702-707. (39) Fujisawa, S., Kadoma, Y., and Masuhara, E. (1986) Effects of (20) Brzozowski, A. M., Pike, A. C., Dauter, Z., Hubbard, R. E., Bonn, photoinitiators for the visible-light resin system on hemolysis of T., Engstrom, O., Ohman, L., Greene, G. L., Gustafsson, J. A., dog erythrocytes and lipid peroxidation of their components. J. and Carlquist, M. (1997) Molecular basis of agonism and antago- Dent. Res. 65, 1186-1190. nism in the oestrogen receptor. Nature 389, 753-758. (40) Seidegard, J., DePierre, J. W., Morgenstern, R., Pilotti, A., and (21) Shiau, A. K., Barstad, D., Loria, P. M., Cheng, L., Kushner, P. Ernster, L. (1981) Induction of drug-metabolizing systems and J., Agard, D. A., and Greene, G. L. (1998) The structural basis of related enzymes with metabolites and structural analogues of estrogen receptor/coactivator recognition and the antagonism of stilbene. Biochim. Biophys. Acta 672,65-78. this interaction by tamoxifen. Cell 95, 927-937. (41) McKim, J. M., Jr., Choudhuri, S., Wilga, P. C., Madan, A., Burns- (22) Shi, L. M., Fang, H., Tong, W., Wu, J., Perkins, R., Blair, R., Naas, L. A., Gallavan, R. H., Mast, R. W., Naas, D. J., Parkinson, Branham, W., and Sheehan, D. (2001) QSAR models using a A., and Meeks, R. G. (1999) Induction of hepatic xenobiotic large diverse set of estrogens. J. Chem. Inf. Comput. Sci. 41 (1), metabolizing enzymes in female Fischer-344 rats following re- 186-195. peated inhalation exposure to decamethylcyclopentasiloxane (D5). (23) Blair, R., Fang, H., Branham, W. S., Hass, B., Dial, S. L., Moland, Toxicol. Sci. 50,10-19. C. L., Tong, W., Shi, L., Perkins, R., and Sheehan, D. M. (2000) (42) Kala, S. V., Lykissa, E. D., Neely, M. W., and Lieberman, M. W. Estrogen Receptor Relative Binding Affinities of 188 Natural and (1998) Low molecular weight silicones are widely distributed Xenochemicals: Structural Diversity of Ligands. Toxicol. Sci. 54, after a single subcutaneous injection in mice. Am. J. Pathol. 152, 138-153. 645-649. (24) Branham, W. S., Dial, S. L., Moland, C. L., Hass, B., Blair, R., (43) Bennett, D. R., Gorzinski, S. J., and LeBeau, J. E. (1972) Fang, H., Shi, L., Tong, W., Perkins, R., and Sheehan, D. M. Structure-activity relationships of oral organosiloxanes on the (2001) and Mycoestrogen Binding to Rat Uterine male reproductive system. Toxicol. Appl. Pharmacol. 21,55-67. Estrogen Receptor. Am. J. Nutr. (in press). (44) Hayden, J. F., and Barlow, S. A. (1972) Structure-activity (25) Fang, H., Tong, W., Perkins, R., Soto, A., Prechtl, N., and relationships of organosiloxanes and the female reproductive Sheehan, D. M. (2000) Quantitative comparison of in vitro assays system. Toxicol. Appl. Pharmacol. 21,68-79. for estrogenic activity. Environ. Health Perspect. 108, 723-729. (45) Guzelian, P. S. (1982) Comparative toxicology of (26) Zacharewski, T. (1998) Identification and assessment of endocrine (kepone) in humans and experimental animals. Annu. Rev. disruptors: limitations of in vivo and in vitro assays. Environ. Pharmacol. Toxicol. 22,89-113. Health Perspect. 106 (Suppl. 2), 577-582. (46) Wilson, N. K., and Zehr, R. D. (1979) Structures of some kepone (27) Meylan, W., and Howard, P. (1995) Atom/Fragment Contribution photoproducts and related chlorinated pentacyclodecanes by Method for Estimating Octanol-Water Partition Coefficients. J. carbon-13 and proton nuclear magnetic resonance. J. Org. Chem. Pharm. Sci. 84,83-92. 44, 1278-1282. (28) Von Angerer, E. (1995) in The Estrogen Receptor As Target For (47) Keasling, H., and Schueler, F. (1950) The relationship between Rational Drug Design (von Angerer, E., Ed.) p 36, R. G. Landes estrogenic action and chemical constitution in a group of azo- Co., Austin, TX. methine derivatives. J. Am. Pharm. Assoc. 39,87-90. (29) Bucourt, R., Vignau, M., and Torelli, V. (1978) New biospecific (48) Abraham, M., Duce, P., and Prior, D. (1989) Hydrogen Bonding. adsorbents for the purification of estradiol receptor. J. Biol. Chem. Part 9. Solute Proton Donor and Proton Acceptor Scales for Use 253, 8221-8228. in Drug Design. J. Chem. Soc., Perkin Trans. 2, 1355-1375. (30) Jordan, V. C., Mittal, S., Gosden, B., Koch, R., and Lieberman, (49) Korach, K. S., Chae, K., Levy, L. A., Duax, W. L., and Sarver, P. M. E. (1985) Structure-activity relationships of estrogens. En- J. (1989) Diethylstilbestrol metabolites and analogs. Stereochem- viron. Health Perspect. 61,97-110. ical probes for the estrogen receptor binding site. J. Biol. Chem. (31) Pike, A., Brzozowski, A., Hubbard, R., Bonn, T., Thorsell, A., 264, 5642-5647. Engstrom, O., Ljunggren, J., Gustafsson, J., and Carlquist, M. (50) Duax, W. L., Griffin, J. F., Weeks, C. M., and Korach, K. S. (1985) (1999) Structure of the ligand-binding domain of oestrogen Molecular conformation, receptor binding, and hormone action receptor â in the presence of a partial agonist and a full of natural and synthetic estrogens and antiestrogens. Environ. antagonist. EMBO J. 18, 4608-4618. Health Perspect. 61, 111-121. (32) Nishirara, T., Nishikawa, J., Kanayama, T., Dakeyama, F., Saito, (51) Nakai, M., Tabira, Y., Asai, D., Yakabe, Y., Shimyozu, T., Noguchi, K., Imagawa, S., Takatori, S., Kitagawa, Y., Hori, S., and Utsumi, M., Takatsuki, M., and Shimohigashi, Y. (1999) Binding charac- H. (2000) Estrogenic Activities of 517 Chemicals by Yeast Two- teristics of dialkyl phthalates for the estrogen receptor. Biochem. Hybrid Assay. J. Health Sci. 46, 282-298. Biophys. Res. Commun. 254, 311-314. (33) Anstead, G. M., Wilson, S. R., and Katzenellenbogen, J. A. (52) Tanenbaum, D. M., Wang, Y., Williams, S. P., and Sigler, P. B. (1989) 2-Arylindenes and 2-arylindenones: Molecular structures (1998) Crystallographic comparison of the estrogen and proges- and considerations in binding orientation of unsymmetrical terone receptor’s ligand binding domains. Proc. Natl. Acad. Sci. nonsteroidal ligands to the estrogen receptor. J. Med. Chem. 32, U.S.A. 95, 5998-6003. 2163-2172. (34) Miksicek, R. J. (1995) Estrogenic flavonoids: structural require- ments for biological activity. Proc. Soc. Exp. Biol. Med. 208, 44-50. TX000208Y